Javier Olaizola Casín’s Post

View profile for Javier Olaizola Casín, graphic

Global Managing Partner Data and Technology Transformation | Let’s create AI with integrity, not bias

The impact of this move is comparable with the creation of Linux. This is a signature moment for the overall GenAI industry. It is a MASSIVE step forward to unlock the value of the open community, and makes GenAI trully accessible and efficient.

View profile for Shobhit Varshney, graphic

VP & Sr. Partner, Americas AI GenAI IoT Lead at IBM [shobhitvarshney.com] 2023 Tech Leader of the Year

What a week for open-source.. another 🎤 mic drop from IBM At the #RedhatSummit, Matt Hicks just announced 🚀 InstructLab to help the AI community collaborate on open-source models with knowledge and skills. Today's open source LLM community has some fundamental problems: ⛔️ each AI engineer creates their own copy of their fine-tuned model, and there is no way for the contribute back to the model or collaborate as a team ⛔️ models are released by providers as major releases months apart, not weekly continuous improvements. eg. despite 15k+ community variants of Llama2, Llama3 was released after 9 months, without a mechanism for others to contribute. ⛔️ it takes a lot for AI engineers to build the right training data sets to fine-tune open source models IBM Research and RedHat worked relentlessly to solve this: ✅ Project InstructLab creates a novel mechanism to reduce the barrier to create synthetic training data, and enable the community to contribute ✅ Taxonomy-based skill and knowledge representation ✅ Synthetic data generation with teacher model ✅ Synthetic data validation with critic model ✅ Skill and knowledge training on top of the student model(s) This drastically reduces the barrier and timeline to fine-tune a model with minimal effort and data. To prove how well it works, look at how IBM applied InstructLab on our granite code models (that we open-sourced yesterday https://lnkd.in/gPz2FPBh) and crushed the competition on benchmarks. Replay of Keynote: https://lnkd.in/g-eTrz5m Project: https://instructlab.ai/ News: https://lnkd.in/eVNctKV6 Explainer: https://lnkd.in/eJ3ZqN9A Paper: https://lnkd.in/e3r25GuT Hats off to the amazing collaboration between IBM Research (Sriram Raghavan, David Cox, Ruchir Puri, Dario Gil & team) and Red Hat (Matt Hicks, Ashesh Badani & team) Mohamad Ali, Blaine Dolph, Kelly Chambliss, Matthew Candy, Glenn Finch, Javier Olaizola Casín, Jesús Mantas -------- 🔔 If you like such content, I encourage you to connect, comment and repost ♻️ Subscribe to my free ‘AI with Whiskey’ newsletter https://lnkd.in/gK4zrSYv -------- #trulyopen #opensource #aiforgood

  • No alternative text description for this image
  • No alternative text description for this image

To view or add a comment, sign in

Explore topics